Emissivity Segmentation Instance Segmentation Dataset and Pre-Trained Model by Lukas Seitz

Images

240 images

Explore Dataset

Introduction

In passive thermography, infrared radiation is used to measure the temperature of surfaces. An important factor thereby is the emissivity of different surface materials. The emissivity determines how well a material emits infrared radiation and therefore specifies the ratio of reflected radiation from the environment to self-emitted radiation. Different material have different emissivities, such that temperature measurements are distorted if the respective emissivities are not taken into account.

Infrared cameras measure the intensity of thermal radiation coming from an object and create infrared (IR) images or thermograms from this. For simplicity, usually only one general emissivity value for the entire image is thereby assumed, ignoring different emissivities. A precise temperature measurement would however require to know the individual emissivity of every pixel of the IR-image and calibrate the temperature values accordingly.

Goal of this project

In order to improve the accuracy of thermal images, a method for temperature calibration is developed in this project. Hereby, an additional image of the same scene is taken by a conventional visible light camera, which is based on red-green-blue (RGB) color channels. This RGB-image is then used to detect and classify different surface materials using a deep segmentation network. Once the materials are known, they can be assigned to their individual emissivity value. In this way, the original thermal image can be corrected on an automatic basis.

Dataset

In order to train the segmentation model, 120 RGB-images of a sample plate were taken. The plate consists of 16 different materials (aluminum, black paint, white paint, thermo foil, hard foam, glimmer, copper corrosion, shiny copper, wood, cork, stainless steel, pvc, granite, marble, teflon, mortar). One corner of the aluminum sample plate is additionally roughened.

Further information about the project can be found here: https://gitlab.lrz.de/ge36bob/emissivity-segmentation

Trained Model API

This project has a trained model available that you can try in your browser and use to get predictions via our Hosted Inference API and other deployment methods.

Cite This Project

If you use this dataset in a research paper, please cite it using the following BibTeX:

@misc{
                            emissivity-segmentation_dataset,
                            title = { Emissivity Segmentation Dataset },
                            type = { Open Source Dataset },
                            author = { Lukas Seitz },
                            howpublished = { \url{ https://universe.roboflow.com/lukas-seitz-qrkko/emissivity-segmentation } },
                            url = { https://universe.roboflow.com/lukas-seitz-qrkko/emissivity-segmentation },
                            journal = { Roboflow Universe },
                            publisher = { Roboflow },
                            year = { 2024 },
                            month = { mar },
                            note = { visited on 2024-05-04 },
                            }